Database Reference
In-Depth Information
To extract the features for data mining, the rough set theory [1] is applied. By us-
ing the rough set theory, a decision rule induction from an attribute value table is
done. The feature extraction algorithm can generate multiple feature sets (reducts).
These feature sets are used for predicting the user's action with the primary decision-
making algorithm and confirmation algorithm. The primary decision-making
algorithm compares the feature values of objects with decision rules. If a matching
criterion is found, the decision rule for action of the speech interactive agent is as-
signed to the specific job. However, the user may not require the specific task to be
performed because of lack of confidence if the user is distracted at that time. Thus,
the confirmation algorithm is applied using speech interaction tools; speech recogni-
tion and text-to-speech. When the user just says ''yes”, the action is performed ac-
cording to the rule of the decision-making algorithm.
Table 2.
Decision rules for the action
Decision rule 1. IF (F1 = 0) THEN (D = 0)
Decision rule 2. IF (F2 = 1) AND (F3 = NOW) THEN (D = N)
Decision rule 3. IF (F4 = 1) AND (F5 = 1) THEN (D = N)
Table 3.
Test sample data
Object No.
F1
F2
F3
F4
F5
D
1,2,3,4,5,6,7
0
X
X
X
X
0
1
1
1
Time
24%
2
1
2
1
1
Time
10%
3
2
3
1
0
X
4%
5
3
4
1
0
X
10%
4
4
5
1
0
X
50%
1
5
6
1
0
X
1%
7
6
7
1
0
X
1%
6
7
We select five features, F1-F5. F1 is the indicator to notify whether the system is
in the sleep mode or not. F2 is the indicator to notify whether the object (application
ID) is one reserved at the scheduled time or not. F3 is the reserved time if the F2 is
set to 1. F4 is the frequency rate when the object is used for some time. F5 is the
priority of that application. Table 2 includes 3 decision rules generated with the rule
extraction algorithm. The decision rules are followed continually when the F1 is just
set to 1. If the matching criterion is met in the next decision rules, decision rule is set
to N, which is the object number to be performed by the speech interactive agent.
Table 3 depicts a sample data set. When F1 is 1, F2 is 1, and F3 of the object 2 is on
time to be executed, the decision rule, D is set to 2. Thus, the object 2 is selected as
the one that can be executed. If F3 does not notify by a scheduled time, the decision
rule,
D
is set to 5 because the object number 5 has the highest priority,
F5=1
.
This proposed method gives the intelligence and automation for user satisfaction.
In addition, sensory fused rules and data fusion rules can be employed easily on the
Search WWH ::
Custom Search